A distributed team cannot hold the same picture in its head.

When a team is distributed, the shared picture lives in whoever happened to be in the conversation. Someone changes how the product behaves and the people writing about it find out later, or never. Someone runs an experiment and the result stays with them. The same decision gets made twice because nobody could find the first one.

The usual answer is more meetings, which is a tax paid weekly forever. We wanted the alternative: one place holding what is currently true, that everyone reads from, including the AI agents doing real work alongside the team.

The brain is a git-tracked knowledge base that agents read from and write to, organised around one keystone rule. Durable knowledge lives as notes. Anything that moves on its own, a metric, a deploy status, a model version, is never written down as a fact. It is stored as a pointer to wherever the live answer lives, and a reading agent fetches it fresh.

That single distinction is what stops a knowledge base rotting. Notes do not go stale because the things that go stale were never put in them.

The effect is that the business stays in step with itself. When the product changes, the corpus changes, and the person writing the next campaign is working from what is true today rather than what was true when they last asked. When marketing learns something from customers, product can see it without being told. Nobody has to remember to pass it on.

The part that makes this different from a well-kept wiki is who does the keeping. The agents are not consuming documentation someone else maintains. They write the daily briefing, take the dated reading of the metrics, report what shipped, and review every note against its own review date. They are bound by the same six rules as the people. Most companies pointing an AI at their internal documents have a read-only relationship with a corpus that is already stale.

The architecture document names its own failure modes, including the unsolved ones. The weekly steward only fires while a particular machine is awake, and that is written down rather than left to be discovered. The current version records that an earlier version described folders which were never built.

Changing how any of it works requires a decision note explaining why, in the same change as the edit. The map cannot drift from the territory without leaving a paper trail.

We do. See Stella was founded in 2023 by Tanuja and Rishabh with a dermatologist co-founder. It is where we work out how an AI-native operation should run before we bring any of it to a client, and it is the reason we can describe this from the inside rather than from a conference talk.

How we implement AI →